生产就绪型 AI：使用 Amazon Bedrock 构建弹性应用程序.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AIM 3306Vadim OmeltchenkoHe/HimSr.AI/ML Solutions ArchitectAmazon Web ServicesEllen HsuShe/HerSr.M

2、anager Amazon BedrockAmazon Web ServicesProduction-Ready AI:Architecting Resilient Apps with Amazon Bedrock 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.IntroductionBuilding blocks of resilient GenAI applications-Model Choice-Scale-Fault tolerance-Security and GovernanceQ&AAgen

3、da 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Lets launch a startup!2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Model ChoiceIntelligence,Performance,CostCustomization optionsEvaluationsOrchestrationBuilding blocks of a resilient applicationProprietary v

4、s Open sourceGeo availability 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.ModelsBedrockAgentCoreSingle tenant architecture 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.ModelsBedrockAgentCoreReplicate for other tenants 2025,Amazon Web Services,Inc.or its a

5、ffiliates.All rights reserved.Launched!(1 week later)2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.The ThrottlingException error 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Lets scale our startup!2025,Amazon Web Services,Inc.or its affiliates.All rights re

6、served.ScaleBuilding blocks of a resilient applicationTPM/RPM needsQuotas and monitoringSync/AsyncCross Region Inference(CRIS)2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.ModelsBedrockCRISAgentCoreModelsBedrockRegion 1Region 2Cross-region inference profiles(CRIS)2025,Amazon Web